Projects / Programmes source: ARIS

Automatic object-oriented land cover classification of optical remote sensing data

Research activity

Code Science Field Subfield
2.17.00  Engineering sciences and technologies  Geodesy   

Code Science Field
T181  Technological sciences  Remote sensing 

Code Science Field
2.07  Engineering and Technology  Environmental engineering  
remote sensing, satellite images, land cover, segmentation, object-oriented classification, automatic pre-processing
Evaluation (rules)
source: COBISS
Researchers (8)
no. Code Name and surname Research area Role Period No. of publicationsNo. of publications
1.  35010  Nataša Đurić  Geodesy  Researcher  2014 - 2017  30 
2.  25640  PhD Žiga Kokalj  Geography  Researcher  2014 - 2017  369 
3.  28658  PhD Aleš Marsetič  Geodesy  Researcher  2014 - 2017  105 
4.  15112  PhD Krištof Oštir  Geodesy  Head  2014 - 2017  589 
5.  25040  Peter Pehani    Technical associate  2014 - 2017  97 
6.  36950  Maja Somrak  Computer science and informatics  Technical associate  2017  28 
7.  35230  MSc Andreja Švab Lenarčič  Geodesy  Researcher  2014 - 2017  25 
8.  20005  PhD Tatjana Veljanovski  Geodesy  Researcher  2014 - 2017  154 
Organisations (2)
no. Code Research organisation City Registration number No. of publicationsNo. of publications
1.  0618  Research Centre of the Slovenian Academy of Sciences and Arts  Ljubljana  5105498000  61,885 
2.  3039  Centre of Excellence for Space Sciences and Technologies  Ljubljana  3665313  744 
Satellite images are an important source of environmental data and have proven to be highly effective in many applications. With an increasing number of remote sensing data users we are witnessing a strong need for simple products that are delivered in near real time and without human intervention. Land cover maps provide an important insight into the state of environment and are of great importance in Earth sciences. In Slovenia, they are mainly produced by visual interpretation of aerial photographs, while in other countries satellite images are used as a primary data source. We have identified three main research challenges in land cover production: under exploitation of satellite imagery, non-optimal use of segmentation algorithms, and non-optimal use of classification algorithms. Although the potential for the use of satellite imagery is well demonstrated, their actual exploitation has not increased in recent years. The main reasons for this, other than the price, are time-consuming and complex pre-processing, and a complicated classification procedure. Due to increase in spatial and spectral resolution of satellite imagery pixel-based classification has been replaced almost entirely by object-based classification. Potential of its main steps – i.e. segmentation and classification – is not yet fully studied, however. The main aim of the proposed project is to develop an automatic procedure for geometric and radiometric correction of satellite images, and production of a land cover map (vector layer) that can be used directly by end-users in their analyses using geographic information systems (GIS). The proposed project is organized in four work packages: ·         WP1: Automatic geometric and radiometric corrections of optical satellite data, ·  WP2: Development and optimisation of the input segmentation image and variables, ·         WP3: Development of non-parametric supervised classification methods and ·         WP4: Integration, automation and optimisation of developed procedures. In WP1 we will develop a robust automatic geometric correction procedure that will be based on rational polynomial coefficients (RPCs). WP2 will be devoted to segmentation with existing algorithms and analysing the influence of different input data characteristics (sensor type, spatial and spectral resolution, computed products). We will also introduce texture into segmentation. In WP3 we will analyse different classification methods, and look for significant attributes for optimal within class separation and training samples determination on a multi-level scale. In WP4 we will integrate an iterative method defined by WP2 and WP3, to automatically classify incoming remote sensing data, pre-processed by a procedure developed in WP1. The project is being proposed by the group that has proven – by end users, founding organizations and with publications – to be effective in various applied and research remote sensing projects. We already have all the necessary hardware, software, and data needed for a successful realization of the project. The project will deal with several important research topics, but it is important to emphasize the potential for application of the results. The extensive analyses that are going to be performed will enable commercial entities – including a potential spin-off – to transfer the obtained knowledge to systems that are currently used by prospective end-users (ez.g. Ministry of Agriculture and the Environment of the Republic of Slovenia). The project is harmonized with the needs and demands of current project calls, published by European Space Agency and European Union. In the dawning of the huge amount of data coming from the Sentinel satellites and the European initiative Copernicus, the project is important not only in strengthening national initiatives in Slovenia, but also in putting Slovenia on the Copernicus roadmap.
Significance for science
The field of automatic object­based classification has been a subject of intensive research, but still far from established solutions. The project dealt with state­of­the­art research and brought new knowledge to the Slovenian and international community, providing benefits to science. The research introduces an innovative, fully automatic process of geometric, radiometric and atmospheric corrections, and semi-automatic land cover map production. According to the review of existing systems, it represents a novelty worldwide and, in case of implementation, a great competitive advantage for (Slovenian) companies operating in the field of geoinformation technologies. Given the experiences in the profession, such procedures are also eagerly awaited by many users. In the first phase, the advancement of the approach was reflected in the fully automatic image pre-processing. Partially already developed and functional set of procedures was complemented, which resulted in an improved accuracy of geometric and radiometric corrections in comparison with commercially available software. The procedure also accelerates the acquisition of data. The introduction of texture as one of the most important inputs for segmentation represents an advanced approach to automatic object detection. The main contribution of the project is a systematic study of the various input data and derivate layers, for example, vegetation indexes, texture, etc. and the selection of the optimal input segmentation image. Our contribution to the know­how in the field was also in the use of multi­level segmentation. Several algorithms were tested in the classification phase, especially those that have not been used for (automatic) detection of specific land cover yet. We thoroughly analysed the resulting segments obtained from the previous phase. Kolmogorov Smirnov test, one of the explored and tested algorithms, was very effective compared to standard methods;, therefore, its integration in the procedure is of high importance. The land cover classes at several levels were defined according to the European guidelines and recommendations for improved comparability of layers of cover in different European countries. A prototype of processing chain for the production of land cover vector layer developed within the framework of the project provides a good starting point for further optimization of high resolution and open-access satellite data classification and transfer to operational use.
Significance for the country
The research performed in this project improved and radically shortened the procedure for the production of land cover layers. This is especially important when these maps should be produced repeatedly (e.g. regular monitoring of a specific event, process or object). Results of the project were envisaged with end-users in mind, for example, the agencies and working groups of the Ministry of Agriculture and the Environment. The main contribution is an improvement of the established slow processes of land use identification and interpretation, consequently lowering the costs in public sector (field work, subsidies, raising the level of readiness for sterner natural and economic conditions in the country etc.). It was, therefore, essential to developing these methods to be transferable to other geographical areas. The developed solutions are used for faster and better work in the companies. Our strong collaboration with Sinergise Ltd., one of the most innovative companies in the field of GIS applications in agriculture, is showing the cooperation between the industry and the research community with solutions for concrete problems. Areas where we already successfully cooperate with companies are continuous observations, monitoring and identification of draught, services for smart agriculture, monitoring and rapid mapping during natural disasters. These activities are exclusively related to remote sensing data, which we mainly acquired from the European Space Agency and other space agencies. Remote sensing enables continuity and repeatability of observations over large areas, which is a prerequisite for the establishment of up-to-date systems in any geographical area. Slovenian and foreign end-users have acknowledged our existing solutions as effective and efficient. The results of the project are useful for representatives of Slovenian and international organisations as they are valuable for spatial planning and thus indirectly applicable in solving social problems. Spatial solutions mentioned in the methodology are important both locally and globally. With the obtained results we can significantly reduce losses or damage in case of ecological catastrophes or natural disasters. In addition, the preventive actions carried out using this methodology can maintain the biodiversity of a given local area in a long term. The fact that access to geographical or spatial information today is at the level of public good is also significant. Therefore good quality, correctly interpreted and updated geographic information is of extreme importance.
Most important scientific results Annual report 2014, 2015, final report
Most important socioeconomically and culturally relevant results Annual report 2014, 2015, final report
Views history